Tests Should Not Be Optional

I am of the opinion that any code that does not have accompanying automated tests should be considered inherently defective. When we write code we have certain notions of suitableness in mind. Tests are a way to express those notions in a way that others can run, and examine, independent of having to delve into our actual implementation. Tests provide a way to communicate about the code we write in regards to what it does, not simply how it does some thing. While those two notions of “what code does,” and “how code does something” seem almost like the same thing, I believe there is value in distinguishing them, and that distinction is why I think code without tests should be considered inherently defective.

What versus How

When we write code we are being prescriptive, we are stating with as much precision as our chosen language allows, or implies, the steps we want the computer to take in completing some work. Our implementation is a judgement about the best way, at a given point in time to accomplish something. This judgement has assumptions baked in that are derived from the context within which the code was written. When we commit to a particular implementation we are making the assertion that the code we are writing ought to work. But, the code we write makes only the most minimal of guarantees, and those guarantees it does make strongly are highly…